Error capturing to a subfolder

[george1421]

I’m currently working through another forum members issue where they are trying to augment their fog server’s storage with a second hard drive. The forum member created a directory under /images. For the sake of discussion lets call it vol1. Then the new hard drive [/dev/sdb1] was mounted to that new directory [/images/vol1]. Then an new image definition was setup to capture to that subdirectory under the /images directory. A capture was scheduled and the capture failed with unable to locate image store.

<I’ll link in the original thread here>

[george1421]

In my dev environment I have a dev fog server setup for another task so I decided to see if I could duplicate the OP’s issue (short answer is I can).

Here is how I setup the experiment.

1. Added a new vmdk to the fog server
2. Used fdisk to create a partition
3. Formatted the partition ext4
4. Create a directory under /images called vol1
5. Created an image definition that looked like this:
   
6. I assigned that image definition to an existing host
7. Scheduled a capture task
8. The image captured correctly to the directory /images/vol1/VolExtnd1 (sweet)
9. I ran the following command mount -t ext4 /dev/sdb1 /images/vol1
10. I schedule a new capture task
11. Capture failed with Unable to locate image store.
12. I reboot the target computer (since I wasn’t able to umount /images/vol1 because of a locked file ??). Note the same capture task was still pending cause I was running in debug mode.
13. I unmounted /dev/sdb1 from /images/vol1
14. When the target computer rebooted it started imaging and ran to completion placing the files in /images/vol1/VolExtnd1.

So for some reason it doesn’t like a mounted hard drive in this location.

At the point of the error this is what was mapped on the FOS engine (interesting that dev wasn’t mounted yet)

from the FOG server this is what the directory structure looks like

[root@devfog images]# df -h
Filesystem               Size  Used Avail Use% Mounted on
/dev/mapper/centos-root   27G  8.8G   18G  34% /
devtmpfs                 1.9G     0  1.9G   0% /dev
tmpfs                    1.9G     0  1.9G   0% /dev/shm
tmpfs                    1.9G  8.6M  1.9G   1% /run
tmpfs                    1.9G     0  1.9G   0% /sys/fs/cgroup
/dev/sda1                497M  172M  326M  35% /boot
tmpfs                    380M     0  380M   0% /run/user/0
/dev/sdb1                 40G   49M   38G   1% /images/vol1

This is what the images directory looked like

[root@devfog images]# ls -la
total 12
drwxrwxrwx.  7 fog  root  100 May 25 14:55 .
dr-xr-xr-x. 20 root root 4096 May 25 14:12 ..
drwxrwxrwx.  3 fog  root   44 May 25 15:07 dev
drwxrwxrwx.  3 fog  root   26 May 18 17:56 drivers
-rwxrwxrwx.  1 fog  root    0 May 16 15:32 .mntcheck
drwxrwxrwx.  2 fog  root   29 May 16 15:32 postdownloadscripts
drwxrwxrwx.  3 fog  root 4096 May 25 13:30 vol1
drwxrwxrwx.  2 fog  root 4096 May 25 13:46 WIN7OEMX64

This is the point of the actual error when running on FOS. Line #76 of fog.download

[Tom Elliott]

1. You can’t unmount a disk/volume if it’s busy. For example, If you’ve mounted a filesystem to /images/vol1, then you cd to /images/vol1, you won’t be able to unmount until you are out of /images/vol1.

2. Capturing a file to a non-mounted directory does not make the file exist in the mounted point.

• For example: you have created directory /images/vol1. You then captured an image to /images/vol1/someimage. From the /images/vol1 point (nothing mounted) the files are all good and dandy. When you mount the filesystem on /images/vol1, files within /images/vol1 do not become automatically available to the volume you mounted on top of it. You might be able to acheive this with the -o bind arguments, though I’m not 100% sure that would do it.

So let’s say /dev/sda1 is the current /images/vol1
Lets say you want /dev/sdb1 to mount to /images/vol1

Without /dev/sdb1 mounted, the file will exist and be happy on /dev/sda1.
Now you mount the /dev/sdb1 on /images/vol1.
The file still exists in /dev/sda1, but /images/vol1 is no longer pointing at /dev/sda1, it’s pointing at /dev/sdb1.

[george1421]

@Tom-Elliott Yeah I got that on the mounting.

/images/vol1 (which points to a directory on /dev/sda1) images correctly.
/images/vol1 --> /dev/sdb1 errors out

From FOS with the mount point in place if I stop the script where it errors out. I can move to /images/vol1 no problem. I can’t touch a file in /images/vol1 because /images on the fog server is shared out read only. There must be some conditional check that is failing above where it works when /images/vol1 is a directory and failing when /images/vol1 is a mountpoint. I’m just not seeing it.